Monitor datastores IO
|Category:||Drivers - Monitor|
We have a ONE 4.2 with two KVM nodes and datastores on NFS.
Actually, the system datastore use a local disk of the nodes and have long delay for deploy and epilog.
We need to enable some cache and move the system datastore back to the NFS with qcow2
Instead of connection to hosts over SSH and run some
nfsiostat, I could be better to have nice graphics in sunstone.
#1 Updated by Ruben S. Montero over 7 years ago
I am not really sure to get this, your request is to include iostat related metrics to the datastore? If so, could you elaborate it a little bit, for example the command and output you are interested in?
You may want to update monitor_ds.sh script to streamline the metrics... (Note that adapting this script is a recommended/good practice)
#2 Updated by EOLE Team over 7 years ago
This will permit to plan the evolution of storage policy
I saw some libvirt check which handle per VM IO stat, this will permit to identify I/O hungry VMs and move them to other datastores.Maybe this request should be split in two:
- monitoring datastore I/O per host
- monitoring VM I/O